- Email: [email protected]
Accuracy versus precision
The Journal of Emergency
Medicme,
Vol. 8, pp, 641-642,
Printed in the USA. Copyright
1990
ACCURACY
6 1990 Pergamon Press plc
VERSUS PRECISION evident in the arterial blood gas (pH 7.43, calculated arterial bicarbonate 24 mEq/L), perhaps because the arterial sample was immediately placed on ice and run as a “stat” specimen by the laboratory. The pseudohyperkalemia (6.9 mEq/L) is a serum artifact due to clotting in the presence of large numbers of white cells (560,000/ mm3) and platelets (1 ,770,000/mm3). This is not seen in vivo or in plasma samples (6,7). Taken in the aggregate, these precise laboratory values, two of them with decimal points reflecting exacting measurement of electrolytes, are as inaccurate and consequently misleading as the timetable of the Long Island Railroad. Abnormal laboratory data that are internally inconsistent or at variance with the clinical picture should prompt a series of questions rather than precipitous actions unless the patient’s condition is so grave that immediate intervention is necessary. In this case, the questions might go something like this: “Why is a patient with a ~0, of 28 mmHg not cyanotic? Would someone with a blood sugar of 15 mg/dL be awake and conversant? With a potassium of nearly 7 mEq/L, why don’t the precordial T-waves reflect hyperkalemia? Why are the calculated (arterial) and measured (venous) bicarbonate values so divergent?” In many instances such as this one the dictum of “Don’t do something . . . just stand there,” although intuitively unappealing to many action-oriented physicians, suggests one approach to the situation that is at least worthy of consideration. This case holds an important lesson for clinicians: That we must be careful not to be seduced by a piece of laboratory data which, simply because it is arrayed in a quantitative format, provides the illusion of objectivity. While certainly more precise, this information may not be nearly so accurate as one’s thoughtful clinical impression.
While imprecision necessarily implies inaccuracy, the inference that precise measurements are necessarily accurate is illogical. Accuracy requires not only that information be precise, but that it also incline toward the truth (1). The casual tendency that we all have to equate accuracy with precision is easily belied by a commute to or from New York City on the Long Island Railroad. Precise times of arrival and departure are posted, but because they bear only a vague relationship to the seemingly random comings and goings of the trains, we could all agree, after only a brief period of observation, that the schedule is inaccurate and therefore of limited value. For the clinician, like the commuter, the distinction between precision and accuracy represents more than semantics. The paper in this issue by Sacchetti et al is a case in point (2). The authors report on a case of ‘‘leukocyte larceny, ’’ a term coined about ten years ago to describe the phenomenon of spurious hypoxemia due to in vitro metabolism of dissolved 0, by large numbers of white cells after the blood has been removed from the intravascular space (3). In this case report, an independent measurement of hemoglobin saturation by oximetry showed that, although the patient’s p0, was below 50 mmHg, the 0, saturation was actually 94%, reflecting adequate delivery of oxygen to the tissues. It is of interest in this case that not only did the arterial blood gas lack accuracy, but fully half of the results of the SMA-6 were also inaccurate. The glucose and venous bicarbonate were falsely depressed, and the potassium was falsely elevated. Each of these individual phenomena have explanations. As the authors point out, ongoing metabolism of O2 and glucose by large numbers of granulocytes explains the hypoxemia and pseudohypoglycemia ( 15 mg/dL) (4,5). The venous acidosis (bicarbonate 6.7 r&q/L) may reflect a lag in performing the SMA-6 that allowed the leukemic cells to create their own hypoxic environment and increase lactic acid production in vitro. This would account for the severe anion gap acidosis that is not
E. John Gallagher, MD Albert Einstein College of Medicine Bronx, New York Editorial Board
0736-4679/90 641
$3.00 + .OO
642
The Journal of Emergency
REFERENCES 1. Colton T. Statistics in medicine. 2. Sacchetti A, Grynn J, Pope spurious hypoxemia confirmed Med. 1990;8:485-90. 3. Fox M, Brody J, Weintraub Leukocyte larceny: a cause of 1979;67:742-6.
Boston: Little, Brown: 1974:38-9. A, Vasso S. Leukocyte larceny: with pulse oximetry. J Emerg W, Szymanski J, O’Donnell C. spurious hypoxemia. Am J Med.
Medicine
Hess CE, Nichole AB, Hunt WB, Suratt PM. Pseudohypoxemia secondary to leukemia and thrombocytosis. N Engl J Med. 1979;30 1: 361-3. Field JB, Williams HE. Artifactual hypoglycemia sssociated with leukemia. N Engl J Med. 1961;265:946-8. Bellevue R, Dosik H, Spergel G, Gussoff BD. Pseudohyperkalemia and extreme leukocytosis. J Lab Clin Med. 1975;85:660-4. Hartman RC, Mellinkoff SM. Relationship of platelets to serum potassium concentration. J Clin Invest. 1955;34:938.
Medicme,
Vol. 8, pp, 641-642,
Printed in the USA. Copyright
1990
ACCURACY
6 1990 Pergamon Press plc
VERSUS PRECISION evident in the arterial blood gas (pH 7.43, calculated arterial bicarbonate 24 mEq/L), perhaps because the arterial sample was immediately placed on ice and run as a “stat” specimen by the laboratory. The pseudohyperkalemia (6.9 mEq/L) is a serum artifact due to clotting in the presence of large numbers of white cells (560,000/ mm3) and platelets (1 ,770,000/mm3). This is not seen in vivo or in plasma samples (6,7). Taken in the aggregate, these precise laboratory values, two of them with decimal points reflecting exacting measurement of electrolytes, are as inaccurate and consequently misleading as the timetable of the Long Island Railroad. Abnormal laboratory data that are internally inconsistent or at variance with the clinical picture should prompt a series of questions rather than precipitous actions unless the patient’s condition is so grave that immediate intervention is necessary. In this case, the questions might go something like this: “Why is a patient with a ~0, of 28 mmHg not cyanotic? Would someone with a blood sugar of 15 mg/dL be awake and conversant? With a potassium of nearly 7 mEq/L, why don’t the precordial T-waves reflect hyperkalemia? Why are the calculated (arterial) and measured (venous) bicarbonate values so divergent?” In many instances such as this one the dictum of “Don’t do something . . . just stand there,” although intuitively unappealing to many action-oriented physicians, suggests one approach to the situation that is at least worthy of consideration. This case holds an important lesson for clinicians: That we must be careful not to be seduced by a piece of laboratory data which, simply because it is arrayed in a quantitative format, provides the illusion of objectivity. While certainly more precise, this information may not be nearly so accurate as one’s thoughtful clinical impression.
While imprecision necessarily implies inaccuracy, the inference that precise measurements are necessarily accurate is illogical. Accuracy requires not only that information be precise, but that it also incline toward the truth (1). The casual tendency that we all have to equate accuracy with precision is easily belied by a commute to or from New York City on the Long Island Railroad. Precise times of arrival and departure are posted, but because they bear only a vague relationship to the seemingly random comings and goings of the trains, we could all agree, after only a brief period of observation, that the schedule is inaccurate and therefore of limited value. For the clinician, like the commuter, the distinction between precision and accuracy represents more than semantics. The paper in this issue by Sacchetti et al is a case in point (2). The authors report on a case of ‘‘leukocyte larceny, ’’ a term coined about ten years ago to describe the phenomenon of spurious hypoxemia due to in vitro metabolism of dissolved 0, by large numbers of white cells after the blood has been removed from the intravascular space (3). In this case report, an independent measurement of hemoglobin saturation by oximetry showed that, although the patient’s p0, was below 50 mmHg, the 0, saturation was actually 94%, reflecting adequate delivery of oxygen to the tissues. It is of interest in this case that not only did the arterial blood gas lack accuracy, but fully half of the results of the SMA-6 were also inaccurate. The glucose and venous bicarbonate were falsely depressed, and the potassium was falsely elevated. Each of these individual phenomena have explanations. As the authors point out, ongoing metabolism of O2 and glucose by large numbers of granulocytes explains the hypoxemia and pseudohypoglycemia ( 15 mg/dL) (4,5). The venous acidosis (bicarbonate 6.7 r&q/L) may reflect a lag in performing the SMA-6 that allowed the leukemic cells to create their own hypoxic environment and increase lactic acid production in vitro. This would account for the severe anion gap acidosis that is not
E. John Gallagher, MD Albert Einstein College of Medicine Bronx, New York Editorial Board
0736-4679/90 641
$3.00 + .OO
642
The Journal of Emergency
REFERENCES 1. Colton T. Statistics in medicine. 2. Sacchetti A, Grynn J, Pope spurious hypoxemia confirmed Med. 1990;8:485-90. 3. Fox M, Brody J, Weintraub Leukocyte larceny: a cause of 1979;67:742-6.
Boston: Little, Brown: 1974:38-9. A, Vasso S. Leukocyte larceny: with pulse oximetry. J Emerg W, Szymanski J, O’Donnell C. spurious hypoxemia. Am J Med.
Medicine
Hess CE, Nichole AB, Hunt WB, Suratt PM. Pseudohypoxemia secondary to leukemia and thrombocytosis. N Engl J Med. 1979;30 1: 361-3. Field JB, Williams HE. Artifactual hypoglycemia sssociated with leukemia. N Engl J Med. 1961;265:946-8. Bellevue R, Dosik H, Spergel G, Gussoff BD. Pseudohyperkalemia and extreme leukocytosis. J Lab Clin Med. 1975;85:660-4. Hartman RC, Mellinkoff SM. Relationship of platelets to serum potassium concentration. J Clin Invest. 1955;34:938.